minimal test case

test.tex

\makeatletter
\def\input@path{{/home/nbecker/interference/}}
\makeatother
\documentclass[english]{beamer}
\usepackage{amsmath}
\usepackage{fontspec}
\usepackage{unicode-math}
\setcounter{secnumdepth}{3}
\setcounter{tocdepth}{3}

\makeatletter

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% LyX specific LaTeX commands.
%% Because html converters don't know tabularnewline
\providecommand{\tabularnewline}{\\}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Textclass specific LaTeX commands.
 % this default might be overridden by plain title style
 \newcommand\makebeamertitle{\frame{\maketitle}}%
 % (ERT) argument for the TOC
 \AtBeginDocument{%
   \let\origtableofcontents=\tableofcontents
   \def\tableofcontents{\@ifnextchar[{\origtableofcontents}{\gobbletableofcontents}}
   \def\gobbletableofcontents#1{\origtableofcontents}
 }

\makeatother

\usepackage{polyglossia}
\setdefaultlanguage{english}
\begin{document}

\begin{frame}{What should we expect (OFDM)?}

\begin{itemize}
\item OFDM approach
\begin{itemize}
\item If interference covers $1/M$ of the carrier spectrum, we loose $1/M$
of the symbols, which is like $1/M$ of the signal power
\item So if we loose 1/10 of the carriers, it's like receiving 9/10 of the
signal.
\item coding is equivalent to combining the symbols \emph{coherently}, so
we should expect to have to boost the signal by $20\log_{10}(0.9)=0.92\text{{dB}}$
\item In general, the loss is 
\[
20\log_{10}(1-1/M)
\]
\end{itemize}
\end{itemize}
\end{frame}
\end{document}